暴露评估中样本采集量的模拟研究

选择暴露评估常用的4种右偏分布,就评估关注的高百分位数估计与采样量的关系进行模拟研究;又以对数正态分布为代表,从分布形态和变异的角度做了细致探讨.结果表明：（1）对右偏分布来说,百分位数越高,准确估计所需的采样容量就越大.而其估计值都随采样量的增大而趋近理论值,精度也随之增大,采样量500时,本文考察的4种右偏分布除P99.9外的其它百分位数都得到了较为准确的估计.（2）估计相同的百分位数,对数正态分布所需的样本容量要比正态分布大得多;而其分布变异越大,所需的采样量也就越大.本研究可为暴露评估中数据的采样调查提供借鉴.

Simulation Study on Sample Size of Exposure Assessment

In the process of exposure assessment,how to determine the sample size to obtain the most economic and effective percentile estimation has seldom been studied.With the benefit of computer simulation,this paper sets out to explore the relationship with 4 positively skewed distributions.And the lognormal distribution has been further selected to study the impact of distribution pattern and variation on percentile estimates.The simulation results showed as follows：（1）Accurately estimating a higher percentile of the positively skewed distribution would require a larger sample size.And a larger sample size always results in a more accurate and more stable estimated percentile.With a sample size of 500,we accurately estimated all the target percentiles of the 4 positively skewed distributions except the result of P99.9.（2）Estimating the same percentile using lognormal distribution required far larger sample size than it needed when using a normal distribution.Moreover,the sample size needed for a lognormal distribution estimation was positively correlated to the magnitude of the variation of the skewed distribution. This research would be very helpful to the exposure assessment in determining the sample size during the survey period.